The Resilience of Berkeley

Analysis Shows Berkeley Students Remained Engaged in Learning in Spring 2020

Berkeley’s students and faculty proved resilient and adept at immediately pivoting to remote instruction in spring 2020, and continued showing academic success the following fall semester, according to an analysis of campus data.

The analysis, funded by the Office of the Executive Vice Chancellor and Provost, is believed to be the first analysis of the impact of emergency remote instruction in higher education due to COVID-19 using institutional data. GSE Professor Zachary Pardos analyzed data from the learning management system from spring 2020 and semesters prior to the worldwide pandemic, as well as enrollment and grade data up through fall 2020.

“The data didn't show widespread horror stories of instructional failures or student disengagement,” said Pardos, the principal investigator of the study. “The data showed there was little degradation in student engagement and learning. Berkeley’s students and faculty showed resilience in the face of the global health crisis.”

In spring 2020, as students and faculty scrambled to adjust to the sudden shift to remote instruction, the study results showed that many faculty and members of the instructional staff appropriately adjusted their class management, including extending deadlines; increasing the amount of feedback and announcements delivered online; and maintaining a timely grading schedule. At the same time, Berkeley took extraordinary steps to increase student support such as technology and financial assistance; additional advising; and increased tele-health; among other efforts.

By the end of fall 2020, when students had completed the next required sequenced course, which was also held remotely, Pardos’s analysis shows no effect of the pandemic on fall letter grades, but a small effect on classes taken as Pass/No-pass in fall.

“Student online engagement spiked in the first two days of remote instruction,” Pardos said, noting that students also increased their online discussion group activity two-fold throughout the semester compared to pre-pandemic spring semesters.

“There was a worry that some students would fall through the cracks, with no connectivity at home, or submitting late assignments. That did happen for some but overall, students remained engaged with their courses, likely in part thanks to the efforts of the campus to mitigate connectivity problems,” he said.

Faculty, too, showed flexibility and creativity in adjusting to remote instruction despite challenges such as learning new technologies; teaching from home or other remote locations; spotty connectivity; among other hurdles, the analysis showed.

“Faculty substantially increased their digital communications with students, and graded assignments at about the same rate,” Pardos said.

When examining the ethnicity of students, the data showed no significant difference in academic success between in-person and remote instruction semesters, noting though that there was a change in grading practices for spring 2020.

In the week after the 2020 spring break, about 15% of Black and Native American students disengaged from classes as compared with white students, but by the end of the semester, all students were equally engaged, Pardos said.

“We’re unsure why there was a dip in engagement of Black and Native American students, and the data doesn’t suggest why. It is certainly worth further study,” he said.

Longer term, the analysis could help the university understand and improve remote instruction but Pardos cautioned that Berkeley, its students, and its faculty made greater than average efforts during the initial months of the pandemic – something that is unsustainable as a typical course of learning.

“There were untold stressors on students and faculty, and we need to look at the other exceptional costs of remote instruction,” he said.

The rapid analysis was possible because the data already existed as part of Berkeley’s learning management system (bCourses on Canvas) and the Student Information System’s enrollment records, as opposed to data gathered specifically for a study through directly surveying students and faculty.

“I find Professor Pardos’s analysis fascinating and incredibly instructive,” said Jenn Stringer, Berkeley’s Associate Vice Chancellor for IT & Chief Information Officer. “This is a huge piece of the puzzle that can help inform, from a data-driven perspective, what learning looked like as we moved to remote instruction.”

While Pardos’s analysis looked at specific courses for a designated time period, AVC Stringer noted that Berkely has much more data from bCourses on Canvas and other campus learning systems that are ripe for analysis.

“I’m a technologist and I am not a pedagogist. I would be really interested to understand what are the questions that come to mind for those who are informed in instructional pedagogy in terms of how we can mine this data for additional insights,” she said.

Postdoctoral Associate, Massachusetts Institute of Technology
PhD, Worcester Polytechnic Institute
BS, Worcester Polytechnic Institute

Featured Publications

Jiang, W., Pardos, Z.A. (2021) Towards Equity and Algorithmic Fairness in Student Grade Prediction. In B. Kuipers, S. Lazar, D. Mulligan, & M. Fourcade (Eds.) Proceedings of the Fourth AAAI/ACM Conference on Artificial Intelligence, Ethics, and Society (AIES). ACM. Pages 608–617.

Shao, E., Guo, S., & Pardos, Z.A. (2021). Degree Planning with PLAN-BERT: Multi-Semester Recommendation Using Future Courses of InterestProceedings of the AAAI Conference on Artificial Intelligence35(17), 14920-14929. 

Pardos, Z.A., Nam, A.J.H. (2020) A university map of course knowledgePLoS ONE 15(9): e0233207. [visualization

Support for this analysis was provided by:

Funding: Office of the EVCP Catherine P. Koshland (and formerly A. Paul Alivisatos)

Critical support from other campus leadership: Jenn Stringer, Associate Vice Chancellor for IT and Chief Information Officer; and Shawna Dark, Chief Academic Technology Officer and Executive Director of Research, Teaching, and Learning

Data preparation support*: Sandeep Markondiah Jayaprakash (RTL); Mithra Bandi (ED&A); Radha Karichedu (ED&A); and Aswan Movva (ED&A)

*Research, Teaching, and Learning (RTL); Enterprise Data & Analytics (ED&A)

Python notebook assistance: Weijie Jiang, graduate student researcher.